Re: CPU usage problems when using KAIO in V7.3? on SINIX.

In article <7uh5qs$ebp$1@news.xmission.com>, "Obnoxio The Clown" <obnoxio@hotmail.com> writes: > > From: mike@inftec.be (Mike Aerts) >> >>The database is located on raw devices in which case Informix >>uses kernel asynchronous I/O by default. >>The problem I try to describe only occurs when I have an SQL-statement >>that needs to read a large number of rows and only find a few matches. >>In this case when I look into what happens with sar we see the %sys >>is abnormally high > > This is not necessarily unexpected, as the CPU VP is probably having to do a > lot of manual comparisons. And because KAIO is so much faster than AIO on > your box, you're seeing a more severe impact. I agree that there is some work at the level of oninit to do some comparisons which acount for about 30% of the user time but the 70% of system time is simply due to the fact the number of times kaiopoll is called. I'm under the impression that when the kaiopoll is called in a non blocking way and when no data is available from the disk I/O they just call kaiopoll again. And this just burns CPU cycles in a Windows way not a Unix way. ;-) In a multi CPU environment this is not such a problem as you can decide to give up one CPU to handle the KAIO calls, although I'm not 100% sure these calls are only used by the first oninit process. In a single CPU environment the usage of Asynchronous Kernel I/O drains too much CPU power from the other client processes IMHO. One supplementary question : Am I the only one encountering this? > > [SNIP] > >>I'm open to all suggestions and requests for further information and tests. > > onstat -p Informixx Dynamic Server Version 7.31.UC2 -- On-Line -- Up 00:06:37 - 76384 Kbytes Profile dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached 90536 88898 93259 2.92 13 8 3 0.00 isamtot open start read write rewrite delete commit rollbk 92546 41 434 91244 0 0 0 0 0 gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs 0 0 0 0 0 0 0 ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes 0 0 0 28.87 64.14 1 4 bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans 9 0 181 0 0 0 0 2 ixda-RA idx-RA da-RA RA-pgsused lchwaits 2 0 0 2 0 This is the profile taken during the test where the query reads every active page from all chunks so the value for %cached is normal. RA-pgsused is less normal but yes its KAIO. > onstat -R Informix Dynamic Server Version 7.31.UC2 -- On-Line -- Up 00:07:11 -- 76384 Kbytes 8 buffer LRU queue pairs priority levels # f/m pair total % of length LOW MED_LOW MED_HIGH HIGH 0 f 876 100.0% 876 0 579 293 4 1 m 0.0% 0 0 0 0 0 2 f 876 100.0% 876 0 577 290 9 3 m 0.0% 0 0 0 0 0 4 f 876 100.0% 876 0 578 291 7 5 m 0.0% 0 0 0 0 0 6 f 876 100.0% 876 0 610 263 3 7 m 0.0% 0 0 0 0 0 8 f 874 100.0% 874 0 602 270 2 9 m 0.0% 0 0 0 0 0 10 f 874 100.0% 874 0 611 258 5 11 m 0.0% 0 0 0 0 0 12 f 874 100.0% 874 0 584 282 8 13 m 0.0% 0 0 0 0 0 14 F 874 100.0% 874 0 595 269 10 15 m 0.0% 0 0 0 0 0 0 dirty, 7000 queued, 7000 total, 8192 hash buckets, 2048 buffer size start clean at 60% (of pair total) dirty, or 525 buffs dirty, stop at 50% 0 priority downgrades, 0 priority upgrades > onstat -F Informix Dynamic Server Version 7.31.UC2 -- On-Line -- Up 00:07:24 -- 76384 Kbytes Fg Writes LRU Writes Chunk Writes 0 0 3 address flusher state data 24314504 0 I 0 = 0X0 243149f0 1 I 0 = 0X0 states: Exit Idle Chunk Lru > > $INFORMIXDIR/etc/$ONCONFIG Configuration File: /opt/lib/informix/etc/rmsanoisa.conf #************************************************************************** # # INFORMIX SOFTWARE, INC. # # Title: onconfig.std # Description: Informix Dynamic Server Configuration Parameters # #************************************************************************** # Root Dbspace Configuration ROOTNAME rootdbs # Root dbspace name ROOTPATH /infor/d01part00 # Path for device containing root dbspace ROOTOFFSET 0 # Offset of root dbspace into device (Kbytes) ROOTSIZE 300000 # Size of root dbspace (Kbytes) # Disk Mirroring Configuration Parameters MIRROR 1 # Mirroring flag (Yes = 1, No = 0) MIRRORPATH /infor/d02part00 # Path for device containing mirrored root MIRROROFFSET 0 # Offset into mirrored device (Kbytes) # Physical Log Configuration PHYSDBS rootdbs # Location (dbspace) of physical log PHYSFILE 20000 # Physical log file size (Kbytes) # Logical Log Configuration LOGFILES 6 # Number of logical log files LOGSIZE 20000 # Logical log size (Kbytes) # Diagnostics MSGPATH /usr/users/informix/rmsano.log CONSOLE /usr/users/informix/rmsano.con # System console message path ALARMPROGRAM /opt/lib/informix/etc/log_full.sh # Alarm program path SYSALARMPROGRAM /opt/lib/informix/etc/evidence.sh # System Alarm program path TBLSPACE_STATS 1 # System Archive Tape Device TAPEDEV /dev/ios0/rstape006chn # Tape device path TAPEBLK 64 # Tape block size (Kbytes) TAPESIZE 4000000 # Maximum amount of data to put on tape (Kbytes) # Log Archive Tape Device LTAPEDEV /dev/ios0/rstape006chn # Tape device path LTAPEBLK 64 # Log tape block size (Kbytes) LTAPESIZE 4000000 # Max amount of data to put on log tape (Kbytes) # Optical STAGEBLOB # Informix Dynamic Server/Optical staging area # System Configuration SERVERNUM 0 # Unique id corresponding to a Dynamic Server instance DBSERVERNAME rmsanoisashm # default database server DBSERVERALIASES rmsanoisatcp # List of alternate dbservernames NETTYPE tlitcp,2,50,NET # Configure poll thread(s) for nettype NETTYPE ipcshm,1,50,CPU # Configure poll thread(s) for nettype DEADLOCK_TIMEOUT 60